Your Right to Repair AI Systems | Rumman Chowdhury | TED

44,322 views ใƒป 2024-06-05

TED


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืชืจื’ื•ื: zeeva livshitz ืขืจื™ื›ื”: aknv tso
00:04
I want to tell you a story
0
4543
1835
ืื ื™ ืจื•ืฆื” ืœืกืคืจ ืœื›ื ืกื™ืคื•ืจ
00:06
about artificial intelligence and farmers.
1
6419
4171
ืขืœ ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื•ื—ืงืœืื™ื.
00:10
Now, what a strange combination, right?
2
10966
3044
ืฉื™ืœื•ื‘ ืžื•ื–ืจ, ื ื›ื•ืŸ?
00:14
Two topics could not sound more different from each other.
3
14010
4088
ืฉื ื™ ื ื•ืฉืื™ื ืฉืœื ื™ื›ืœื• ืœื”ื™ืฉืžืข ืฉื•ื ื™ื ื™ื•ืชืจ ื–ื” ืžื–ื”.
00:18
But did you know that modern farming actually involves a lot of technology?
4
18431
5047
ืื‘ืœ ื”ืื ื™ื“ืขืชื ืฉื”ื—ืงืœืื•ืช ื”ืžื•ื“ืจื ื™ืช ื›ืจื•ื›ื” ืœืžืขืฉื” ื‘ื”ืจื‘ื” ื˜ื›ื ื•ืœื•ื’ื™ื”?
00:23
So computer vision is used to predict crop yields.
5
23478
3796
ืจืื™ื™ืช ืžื—ืฉื‘ ืžืฉืžืฉืช ืœื—ื™ื–ื•ื™ ื’ื•ื“ืœ ื”ื™ื‘ื•ืœื™ื,
00:27
And artificial intelligence is used to find,
6
27315
2836
ื•ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืžืฉืžืฉืช ืœืื™ืชื•ืจ, ื–ื™ื”ื•ื™ ื•ื—ื™ืกื•ืœ ื—ืจืงื™ื.
00:30
identify and get rid of insects.
7
30151
2878
00:33
Predictive analytics helps figure out extreme weather conditions
8
33071
4129
ื ื™ืชื•ื— ื—ื™ื–ื•ื™ ืขื•ื–ืจ ืœื”ื‘ื™ืŸ ืชื ืื™ ืžื–ื’ ืื•ื•ื™ืจ ืงื™ืฆื•ื ื™ื™ื
00:37
like drought or hurricanes.
9
37242
1960
ื›ืžื• ื‘ืฆื•ืจื•ืช ืื• ื”ื•ืจื™ืงื ื™ื.
00:39
But this technology is also alienating to farmers.
10
39744
4547
ืื‘ืœ ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ื”ื–ื• ื’ื ืžื“ื™ืจื” ื—ืงืœืื™ื,
00:44
And this all came to a head in 2017
11
44291
3003
ื•ื›ืœ ื–ื” ื”ื’ื™ืข ืœืฉื™ื ื‘ืฉื ืช 2017
00:47
with the tractor company John Deere when they introduced smart tractors.
12
47335
4922
ื›ืฉื—ื‘ืจืช ื”ื˜ืจืงื˜ื•ืจื™ื โ€œื’โ€™ื•ืŸ ื“ื™ืจโ€ ื”ืฆื™ื’ื” ื˜ืจืงื˜ื•ืจื™ื ื—ื›ืžื™ื.
00:52
So before then, if a farmer's tractor broke,
13
52299
2961
ืœืคื ื™ ื›ืŸ, ืื ื”ื˜ืจืงื˜ื•ืจ ืฉืœ ื”ื—ืงืœืื™ ื”ืชืงืœืงืœ,
00:55
they could just repair it themselves or take it to a mechanic.
14
55302
3920
ื”ื•ื ื™ื›ื•ืœ ื”ื™ื” ืคืฉื•ื˜ ืœืชืงืŸ ืื•ืชื• ื‘ืขืฆืžื• ืื• ืœืงื—ืช ืื•ืชื• ืœืžื›ื•ื ืื™.
00:59
Well, the company actually made it illegal
15
59222
2961
ื•ื‘ื›ืŸ, ื”ื—ื‘ืจื” ืงื‘ืขื” ืฉื–ื” ื‘ืœืชื™-ื—ื•ืงื™
01:02
for farmers to fix their own equipment.
16
62225
2586
ืฉื—ืงืœืื™ื ื™ื•ื›ืœื• ืœืชืงืŸ ืืช ื”ืฆื™ื•ื“ ืฉืœื”ื.
01:04
You had to use a licensed technician
17
64811
2836
ืขืœื™ื”ื ืœื”ื™ืขื–ืจ ื‘ื˜ื›ื ืื™ ืžื•ืจืฉื”
01:07
and farmers would have to wait for weeks
18
67689
2628
ื•ื”ื—ืงืœืื™ื ื™ื™ืืœืฆื• ืœื—ื›ื•ืช ืฉื‘ื•ืขื•ืช
01:10
while their crops rot and pests took over.
19
70317
3044
ื•ื‘ื™ื ืชื™ื™ื, ื™ื‘ื•ืœื™ื”ื ื™ื™ืจืงื‘ื• ื•ื”ืžื–ื™ืงื™ื ื™ืฉืชืœื˜ื•.
01:14
So they took matters into their own hands.
20
74279
3170
ืื– ื”ื ืœืงื—ื• ืืช ื”ืขื ื™ื™ื ื™ื ืœื™ื“ื™ื”ื.
01:17
Some of them learned to program,
21
77490
1544
ื—ืœืงื ืœืžื“ื• ืœืชื›ื ืช,
ื•ื”ื ืขื‘ื“ื• ืขื ืคืฆื—ื ื™ื ื•ื™ืฆืจื• ื˜ืœืื™ื ื›ื“ื™ ืœืชืงืŸ ืืช ื”ืžืขืจื›ื•ืช ืฉืœื”ื.
01:19
and they worked with hackers to create patches to repair their own systems.
22
79075
4922
01:23
In 2022,
23
83997
1418
ื‘ืฉื ืช 2022,
01:25
at one of the largest hacker conferences in the world, DEFCON,
24
85457
3712
ื‘ืื—ื“ ืžื›ื ืกื™ ื”ืคืฆื—ื ื™ื ื”ื’ื“ื•ืœื™ื ื‘ืขื•ืœื, โ€œื“ืคืงื•ืŸโ€œ,
01:29
a hacker named Sick Codes and his team
25
89210
2586
ืคืฆื—ืŸ ื‘ืฉื โ€œืกื™ืง ืงื•ื“ืกโ€ ื•ื”ืฆื•ื•ืช ืฉืœื•
01:31
showed everybody how to break into a John Deere tractor,
26
91838
3212
ื”ืจืื” ืœื›ื•ืœื ื›ื™ืฆื“ ืœืคืจื•ืฅ ืœื˜ืจืงื˜ื•ืจ ืฉืœ ื’โ€™ื•ืŸ ื“ื™ืจ,
01:35
showing that, first of all, the technology was vulnerable,
27
95091
3629
ื•ื‘ื›ืš ื”ืจืื” ืฉืงื•ื“ื ื›ืœ, ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ืคื’ื™ืขื”,
01:38
but also that you can and should own your own equipment.
28
98762
4337
ืื‘ืœ ื’ื ืฉืื“ื ื™ื›ื•ืœ ื•ืฆืจื™ืš ืœื”ื™ื•ืช ืื—ืจืื™ ืœืฆื™ื•ื“ ืฉืœื•.
01:43
To be clear, this is illegal,
29
103683
2836
ืฉื™ื”ื™ื” ื‘ืจื•ืจ, ื–ื” ืœื ื—ื•ืงื™,
01:46
but there are people trying to change that.
30
106561
2795
ืื‘ืœ ื™ืฉ ืื ืฉื™ื ืฉืžื ืกื™ื ืœืฉื ื•ืช ืืช ื–ื”.
01:49
Now that movement is called the โ€œright to repair.โ€
31
109689
3796
ื”ืชื ื•ืขื” ื”ื–ื• ื ืงืจืืช ื”ื™ื•ื โ€œื–ื›ื•ืช ืœืชื™ืงื•ืŸโ€œ.
01:53
The right to repair goes something like this.
32
113526
2169
ื–ื›ื•ืช ืœืชื™ืงื•ืŸ ืื•ืžืจืช ื‘ืขืจืš ื›ื›ื”:
01:55
If you own a piece of technology,
33
115737
1668
ืื ืืชื ื‘ืขืœื™ื ืฉืœ ืžืฉื”ื• ื˜ื›ื ื•ืœื•ื’ื™ --
01:57
it could be a tractor, a smart toothbrush,
34
117405
2544
ื˜ืจืงื˜ื•ืจ, ืžื‘ืจืฉืช ืฉื™ื ื™ื™ื ื—ื›ืžื”,
01:59
a washing machine,
35
119991
1377
ืžื›ื•ื ืช ื›ื‘ื™ืกื”,
02:01
you should have the right to repair it if it breaks.
36
121368
3086
ืืžื•ืจื” ืœื”ื™ื•ืช ืœื›ื ื”ื–ื›ื•ืช ืœืชืงืŸ ืื•ืชื ืื ื”ื ืžืชืงืœืงืœื™ื.
02:05
So why am I telling you this story?
37
125246
2253
ืœืžื” ืื ื™ ืžืกืคืจืช ืœื›ื ืืช ื”ืกื™ืคื•ืจ ื”ื–ื”?
02:08
The right to repair needs to extend to artificial intelligence.
38
128541
5214
ืืช ื”ื–ื›ื•ืช ืœืชื™ืงื•ืŸ ื™ืฉ ืœื”ื—ื™ืœ ื’ื ืขืœ ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช.
02:14
Now it seems like every week
39
134214
2127
ื ืจืื” ืฉื‘ื›ืœ ืฉื‘ื•ืข ื•ืฉื‘ื•ืข
02:16
there is a new and mind-blowing innovation in AI.
40
136383
3336
ืžื•ืคื™ืข ื‘ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืื™ื–ื” ื—ื™ื“ื•ืฉ ื—ื“ืฉ ื•ืžืจืชืง.
02:19
But did you know that public confidence is actually declining?
41
139719
4880
ืื‘ืœ ื”ืื ื™ื“ืขืชื ืฉืืžื•ืŸ ื”ืฆื™ื‘ื•ืจ ื“ื•ื•ืงื ื™ื•ืจื“?
02:24
A recent Pew poll showed that more Americans are concerned
42
144599
4963
ืกืงืจ โ€œืคื™ื•โ€ ืฉื ืขืจืš ืœืื—ืจื•ื ื” ื”ืจืื”
ืฉื”ืืžืจื™ืงื ื™ื ืžื•ื“ืื’ื™ื ื™ื•ืชืจ ืžืืฉืจ ื ืœื”ื‘ื™ื ืžื”ื˜ื›ื ื•ืœื•ื’ื™ื”.
02:29
than they are excited about the technology.
43
149604
2503
02:32
This is echoed throughout the world.
44
152148
2211
ื–ื” ื”ื“ื”ื“ ื‘ืขื•ืœื ื›ื•ืœื•.
02:34
The World Risk Poll shows
45
154401
1418
โ€œืกืงืจ ื”ืกื™ื›ื•ืŸ ื”ืขื•ืœืžื™โ€ ืžืจืื”
02:35
that respondents from Central and South America and Africa
46
155860
3462
ืฉื”ืขื•ื ื™ื ืขืœื™ื• ืžืžืจื›ื– ื•ื“ืจื•ื ืืžืจื™ืงื” ื•ืžืืคืจื™ืงื”
02:39
all said that they felt AI would lead to more harm than good for their people.
47
159322
6090
ืืžืจื• ื›ื•ืœื ืฉืœื“ืขืชื ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืชื‘ื™ื ืœืื ืฉื™ื ื™ื•ืชืจ ื ื–ืง ืžืืฉืจ ืชื•ืขืœืช.
02:46
As a social scientist and an AI developer,
48
166287
2503
ื›ืžื“ืขื ื™ืช ื—ื‘ืจืชื™ืช ื•ื›ืžืคืชื—ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช,
02:48
this frustrates me.
49
168790
1585
ื–ื” ืžืชืกื›ืœ ืื•ืชื™.
02:50
I'm a tech optimist
50
170417
1293
ืื ื™ ืื•ืคื˜ื™ืžื™ืกื˜ื™ืช ื˜ื›ื ื•ืœื•ื’ื™ืช
02:51
because I truly believe this technology can lead to good.
51
171751
4338
ื›ื™ ืื ื™ ื‘ืืžืช ืžืืžื™ื ื” ืฉื”ื˜ื›ื ื•ืœื•ื’ื™ื” ื”ื–ื• ื™ื›ื•ืœื” ืœื”ื•ื‘ื™ืœ ืœื˜ื•ื‘.
02:56
So what's the disconnect?
52
176464
1919
ืื– ืื™ืคื” ื”ื ืชืง?
02:58
Well, I've talked to hundreds of people over the last few years.
53
178758
3754
ื•ื‘ื›ืŸ, ืฉื•ื—ื—ืชื™ ื‘ืฉื ื™ื ื”ืื—ืจื•ื ื•ืช ืขื ืžืื•ืช ืื ืฉื™ื.
03:02
Architects and scientists, journalists and photographers,
54
182554
3670
ืื“ืจื™ื›ืœื™ื ื•ืžื“ืขื ื™ื, ืขื™ืชื•ื ืื™ื ื•ืฆืœืžื™ื,
03:06
ride-share drivers and doctors,
55
186224
1710
ื ื”ื’ื™ ืฉื™ืชื•ืฃ-ื ืกื™ืขื•ืช ื•ืจื•ืคืื™ื,
03:07
and they all say the same thing.
56
187976
2961
ื•ื›ื•ืœื ืื•ืžืจื™ื ืืช ืื•ืชื• ื”ื“ื‘ืจ:
03:12
People feel like an afterthought.
57
192272
3795
ืื ืฉื™ื ืžืจื’ื™ืฉื™ื ืฉืœื ืœืงื—ื• ืื•ืชื ื‘ื—ืฉื‘ื•ืŸ.
03:17
They all know that their data is harvested often without their permission
58
197485
4338
ื›ื•ืœื ื™ื•ื“ืขื™ื ืฉื”ื ืชื•ื ื™ื ืฉืœื”ื ื ืืกืคื™ื ืœืขืชื™ื ืงืจื•ื‘ื•ืช ืœืœื ืจืฉื•ืชื
03:21
to create these sophisticated systems.
59
201823
2461
ื›ื“ื™ ืœื™ืฆื•ืจ ืžืขืจื›ื•ืช ืžืชื•ื—ื›ืžื•ืช ืืœื”.
03:24
They know that these systems are determining their life opportunities.
60
204325
4171
ื”ื ื™ื•ื“ืขื™ื ืฉืžืขืจื›ื•ืช ืืœื” ืงื•ื‘ืขื•ืช ืืช ื”ื”ื–ื“ืžื ื•ื™ื•ืช ืฉืœื”ื ื‘ื—ื™ื™ื ืฉืœื”ื.
03:28
They also know that nobody ever bothered to ask them
61
208496
3545
ื”ื ื’ื ื™ื•ื“ืขื™ื ืฉืื™ืฉ ืžืขื•ืœื ืœื ื˜ืจื— ืœืฉืื•ืœ ืื•ืชื
03:32
how the system should be built,
62
212083
1502
ืื™ืš ื™ืฉ ืœื‘ื ื•ืช ืืช ื”ืžืขืจื›ืช,
03:33
and they certainly have no idea where to go if something goes wrong.
63
213585
5964
ื•ื‘ื•ื•ื“ืื™ ืฉืื™ืŸ ืœื”ื ืžื•ืฉื’ ืœืืŸ ืœืคื ื•ืช ืื ืžืฉื”ื• ืžืฉืชื‘ืฉ.
03:40
We may not own AI systems,
64
220383
2336
ืื ื• ืื•ืœื™ ืœื ื”ื‘ืขืœื™ื ืฉืœ ืžืขืจื›ื•ืช ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช,
03:42
but they are slowly dominating our lives.
65
222761
2794
ืื‘ืœ ื”ืŸ ืžืฉืชืœื˜ื•ืช ืื˜-ืื˜ ืขืœ ื—ื™ื™ื ื•.
03:45
We need a better feedback loop
66
225597
1501
ืื ื• ื–ืงื•ืงื™ื ืœืœื•ืœืืช ืžืฉื•ื‘ ื˜ื•ื‘ื” ื™ื•ืชืจ
03:47
between the people who are making these systems,
67
227140
2961
ื‘ื™ืŸ ื”ืื ืฉื™ื ืฉืžื™ื™ืฆืจื™ื ืืช ื”ืžืขืจื›ื•ืช ื”ืœืœื•,
03:50
and the people who are best determined to tell us
68
230101
3337
ืœื‘ื™ืŸ ื”ืื ืฉื™ื ืฉื™ื›ื•ืœื™ื ื”ื›ื™ ื˜ื•ื‘ ืœืกืคืจ ืœื ื•
03:53
how these AI systems should interact in their world.
69
233480
3628
ืื™ืš ืžืขืจื›ื•ืช ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ื”ืืœื” ืืžื•ืจื•ืช ืœื ื”ืœ ื™ื—ืกื™-ื’ื•ืžืœื™ืŸ ืขื ืขื•ืœืžื.
03:57
One step towards this is a process called red teaming.
70
237609
3837
ืฆืขื“ ืื—ื“ ื‘ื›ื™ื•ื•ืŸ ื–ื” ื”ื•ื ืชื”ืœื™ืš ื‘ืฉื โ€œืฆื•ื•ืช ืื“ื•ืโ€œ.
04:01
Now, red teaming is a practice that was started in the military,
71
241821
3003
โ€œืฆื•ื•ืช ืื“ื•ืโ€ ื”ื•ื ืชืจื’ื•ืœืช ืฉืžืงื•ืจื” ื‘ืฆื‘ื,
04:04
and it's used in cybersecurity.
72
244866
1919
ื•ื”ื•ื ืžืฉืžืฉ ื‘ืื‘ื˜ื—ืช ืกื™ื™ื‘ืจ.
04:06
In a traditional red-teaming exercise,
73
246826
2628
ื‘ืชืจื’ื™ืœ ืžืกื•ืจืชื™ ืฉืœ ืฆื•ื•ืช ืื“ื•ื,
04:09
external experts are brought in to break into a system,
74
249454
3920
ืžื•ื‘ืื™ื ืžื•ืžื—ื™ื ื—ื™ืฆื•ื ื™ื™ื ื›ื“ื™ ืฉื™ืคืจืฆื• ืœืžืขืจื›ืช,
04:13
sort of like what Sick Codes did with tractors, but legal.
75
253374
4338
ื‘ืขืจืš ื›ืžื• ืžื” ืฉโ€œืกื™ืง ืงื•ื“ืกโ€ ืขืฉื” ืขื ื˜ืจืงื˜ื•ืจื™ื, ืื‘ืœ ื‘ืื•ืคืŸ ื—ื•ืงื™.
04:17
So red teaming acts as a way of testing your defenses
76
257712
3670
ืื– ืฆื•ื•ืช ืื“ื•ื ืžืฉืžืฉ ื›ืืžืฆืขื™ ืœื‘ื—ื•ืŸ ืืช ื”ื”ื’ื ื•ืช ืฉืœื›ื
04:21
and when you can figure out where something will go wrong,
77
261424
3545
ื•ื›ืฉืืชื ื™ื›ื•ืœื™ื ืœื’ืœื•ืช ืื™ืคื” ืžืฉื”ื• ื™ืฉืชื‘ืฉ,
04:25
you can figure out how to fix it.
78
265011
3003
ืชื•ื›ืœื• ืœืžืฆื•ื ืื™ืš ืœืชืงืŸ ืืช ื–ื”.
04:28
But when AI systems go rogue,
79
268056
2377
ืื‘ืœ ื›ืฉืžืขืจื›ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื™ื•ืฆืื•ืช ืžืฉืœื™ื˜ื”,
04:30
it's more than just a hacker breaking in.
80
270433
2920
ื–ื” ื™ื•ืชืจ ืžืกืชื ืคืจื™ืฆื” ืฉืœ ืคืฆื—ืŸ.
04:33
The model could malfunction or misrepresent reality.
81
273394
3754
ื”ืžื•ื“ืœ ืขืœื•ืœ ืœื”ืชืงืœืงืœ ืื• ืœื™ื™ืฆื’ ื‘ืื•ืคืŸ ืฉื’ื•ื™ ืืช ื”ืžืฆื™ืื•ืช.
04:37
So, for example, not too long ago,
82
277190
2044
ื›ืš, ืœืžืฉืœ, ืœืคื ื™ ื–ืžืŸ ืœื ืจื‘,
04:39
we saw an AI system attempting diversity
83
279234
2627
ืจืื™ื ื• ืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืฉืžื ืกื” ืœื™ืฆื•ืจ ืžื’ื•ื•ืŸ
04:41
by showing historically inaccurate photos.
84
281903
3503
ืขืœ ื™ื“ื™ ื”ืฆื’ืช ืชืžื•ื ื•ืช ืœื-ืžื“ื•ื™ืงื•ืช ืžื‘ื—ื™ื ื” ื”ื™ืกื˜ื•ืจื™ืช.
04:45
Anybody with a basic understanding of Western history
85
285406
2670
ื›ืœ ืื“ื ื‘ืขืœ ื”ื‘ื ื” ื‘ืกื™ืกื™ืช ื‘ื”ื™ืกื˜ื•ืจื™ื” ื”ืžืขืจื‘ื™ืช
ื™ื›ื•ืœ ื”ื™ื” ืœื•ืžืจ ืœื›ื ืฉืœื ื”ืื‘ื•ืช ื”ืžื™ื™ืกื“ื™ื
04:48
could have told you that neither the Founding Fathers
86
288076
2502
04:50
nor Nazi-era soldiers would have been Black.
87
290620
2169
ื•ืœื ื”ื—ื™ื™ืœื™ื ื”ื ืืฆื™ื™ื ื”ื™ื• ืฉื—ื•ืจื™ื.
04:54
In that case, who qualifies as an expert?
88
294123
3629
ื‘ืžืงืจื” ื›ื–ื”, ืžื™ื”ื• ื”ืžื•ืžื—ื” ื”ืžื•ืกืžืš?
04:58
You.
89
298670
1168
ืืชื.
05:00
I'm working with thousands of people all around the world
90
300380
2836
ืื ื™ ืขื•ื‘ื“ืช ืขื ืืœืคื™ ืื ืฉื™ื ื‘ืจื—ื‘ื™ ื”ืขื•ืœื
05:03
on large and small red-teaming exercises,
91
303258
2252
ื‘ืชืจื’ื™ืœื™ ืฆื•ื•ืช ืื“ื•ื ื’ื“ื•ืœื™ื ื•ืงื˜ื ื™ื,
05:05
and through them we found and fixed mistakes in AI models.
92
305552
4629
ื•ื”ื•ื“ื•ืช ืœื”ื ืžืฆืื ื• ื•ืชื™ืงื ื• ื˜ืขื•ื™ื•ืช ื‘ืžื•ื“ืœื™ื ืฉืœ ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช.
05:10
We also work with some of the biggest tech companies in the world:
93
310181
3587
ืื ื• ืขื•ื‘ื“ื™ื ื’ื ืขื ื›ืžื” ืžื—ื‘ืจื•ืช ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ื”ื’ื“ื•ืœื•ืช ื‘ืขื•ืœื:
05:13
OpenAI, Meta, Anthropic, Google.
94
313810
2753
โ€œืื•ืคืŸ ืื™ื™-ืื™ื™โ€œ, โ€œืžื˜ืโ€œ, โ€œืื ื˜ืจื•ืคื™ืงโ€œ, โ€œื’ื•ื’ืœโ€œ.
05:16
And through this, we've made models work better for more people.
95
316604
4630
ื•ื›ืš ื’ืจืžื ื• ืœืžื•ื“ืœื™ื ืœืขื‘ื•ื“ ื˜ื•ื‘ ื™ื•ืชืจ ืœืžืขืŸ ืจื‘ื™ื ื™ื•ืชืจ.
05:22
Here's a bit of what we've learned.
96
322277
2168
ื”ื ื” ืงืฆืช ืžืžื” ืฉืœืžื“ื ื•.
05:24
We partnered with the Royal Society in London to do a scientific,
97
324988
3670
ืฉื™ืชืคื ื• ืคืขื•ืœื” ืขื โ€œื”ื—ื‘ืจื” ื”ืžืœื›ื•ืชื™ืชโ€ ื‘ืœื•ื ื“ื•ืŸ
ื‘ืขืจื™ื›ืช ืื™ืจื•ืข ืกื‘ื™ื‘ ืžื™ื“ืข ืฉื’ื•ื™ ื•ืžืกื•ืœืฃ ื™ื—ื“ ืขื ืžื“ืขื ื™ ืžื—ืœื•ืช.
05:28
mis- and disinformation event with disease scientists.
98
328658
3545
05:32
What these scientists found
99
332203
1335
ืžื“ืขื ื™ื ืืœื” ืžืฆืื•
05:33
is that AI models actually had a lot of protections
100
333580
2794
ืฉืœืžื•ื“ืœื™ื ืฉืœ ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื™ืฉ ืœืžืขืฉื” ื”ื’ื ื•ืช ืจื‘ื•ืช
05:36
against COVID misinformation.
101
336416
2294
ืžืคื ื™ ืžื™ื“ืข ืฉื’ื•ื™ ืขืœ ื”ืงื•ืจื•ื ื”.
05:38
But for other diseases like measles, mumps and the flu,
102
338751
3546
ืื‘ืœ ืœื’ื‘ื™ ืžื—ืœื•ืช ืื—ืจื•ืช ื›ืžื• ื—ืฆื‘ืช, ื—ื–ืจืช ื•ืฉืคืขืช,
05:42
the same protections didn't apply.
103
342297
2460
ืœื ื—ืœื• ืื•ืชืŸ ื”ื’ื ื•ืช.
05:44
We reported these changes,
104
344757
1293
ื“ื™ื•ื•ื—ื ื• ืขืœ ื”ืฉื™ื ื•ื™ื™ื ื”ืืœื”,
05:46
theyโ€™re fixed and now we are all better protected
105
346050
3170
ื”ื ืชื•ืงื ื• ื•ืขื›ืฉื™ื• ื›ื•ืœื ื• ืžื•ื’ื ื™ื ื˜ื•ื‘ ื™ื•ืชืจ
05:49
against scientific mis- and disinformation.
106
349220
2670
ืžืคื ื™ ืžื™ื“ืข ืžื“ืขื™ ืฉื’ื•ื™ ื•ืžืกื•ืœืฃ.
05:52
We did a really similar exercise with architects at Autodesk University,
107
352724
4838
ืขืฉื™ื ื• ืชืจื’ื™ืœ ืžืžืฉ ื“ื•ืžื” ืขื ืื“ืจื™ื›ืœื™ื ื‘ืื•ื ื™ื‘ืจืกื™ื˜ืช โ€œืื•ื˜ื•ื“ืกืงโ€œ,
05:57
and we asked them a simple question:
108
357562
1877
ื•ืฉืืœื ื• ืื•ืชื ืฉืืœื” ืคืฉื•ื˜ื”:
05:59
Will AI put them out of a job?
109
359439
2961
ื”ืื ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืชื”ืคื•ืš ืื•ืชื ืœืžื•ื‘ื˜ืœื™ื?
06:02
Or more specifically,
110
362442
2169
ืื• ืœื™ืชืจ ื“ื™ื•ืง,
06:04
could they imagine a modern AI system
111
364611
2043
ื”ืื ื”ื ื™ื›ื•ืœื™ื ืœื“ืžื™ื™ืŸ ืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืžื•ื“ืจื ื™ืช
06:06
that would be able to design the specs of a modern art museum?
112
366696
3670
ืฉืชื•ื›ืœ ืœืขืฆื‘ ื‘ืคืจื•ื˜ืจื•ื˜ ืžื•ื–ื™ืื•ืŸ ืœืืžื ื•ืช ืžื•ื“ืจื ื™ืช?
06:10
The answer, resoundingly, was no.
113
370408
3962
ื”ืชืฉื•ื‘ื”, ืคื” ืื—ื“, ื”ื™ืชื” โ€œืœืโ€œ.
06:14
Here's why, architects do more than just draw buildings.
114
374787
3629
ื›ื™ ืื“ืจื™ื›ืœื™ื ืขื•ืฉื™ื ื™ื•ืชืจ ืžืืฉืจ ืกืชื ืœืฆื™ื™ืจ ื‘ื ื™ื™ื ื™ื.
06:18
They have to understand physics and material science.
115
378458
3336
ืขืœื™ื”ื ืœื”ื‘ื™ืŸ ื‘ืคื™ื–ื™ืงื” ื•ื‘ืžื“ืขื™ ื”ื—ื•ืžืจ.
06:21
They have to know building codes,
116
381836
1585
ืขืœื™ื”ื ืœื”ื›ื™ืจ ืืช ื—ื•ืงื™ ื”ื‘ื ื™ื™ื”,
06:23
and they have to do that
117
383463
1251
ื•ืขืœื™ื”ื ืœืขืฉื•ืช ื–ืืช
06:24
while making something that evokes emotion.
118
384756
2794
ื‘ืขื•ื“ื ื™ื•ืฆืจื™ื ืžืฉื”ื• ืฉืžืขื•ืจืจ ืจื’ืฉ.
06:28
What the architects wanted was an AI system
119
388426
2169
ื”ืื“ืจื™ื›ืœื™ื ืจืฆื• ื‘ืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช
06:30
that interacted with them, that would give them feedback,
120
390595
2753
ืฉืชืขื‘ื•ื“ ื™ื—ื“ ืื™ืชื, ืฉืชืชืŸ ืœื”ื ืžืฉื•ื‘,
06:33
maybe proactively offer design recommendations.
121
393389
2753
ืฉืื•ืœื™ ืชื™ื–ื•ื ื”ืžืœืฆื•ืช ืขื™ืฆื•ื‘.
06:36
And today's AI systems, not quite there yet.
122
396142
3670
ื•ืžืขืจื›ื•ืช ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืฉืœ ื™ืžื™ื ื• ืขื“ื™ื™ืŸ ืœื ืœื’ืžืจื™ ืฉื.
06:39
But those are technical problems.
123
399854
1752
ืื‘ืœ ืืœื” ื‘ืขื™ื•ืช ื˜ื›ื ื™ื•ืช.
06:41
People building AI are incredibly smart,
124
401648
2043
ื”ืื ืฉื™ื ืฉื‘ื•ื ื™ื ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื”ื ื—ื›ืžื™ื ืœื”ืคืœื™ื,
06:43
and maybe they could solve all that in a few years.
125
403733
2753
ื•ืื•ืœื™ ื”ื ื™ื•ื›ืœื• ืœืคืชื•ืจ ืืช ื›ืœ ื–ื” ื‘ืขื•ื“ ื›ืžื” ืฉื ื™ื.
06:46
But that wasn't their biggest concern.
126
406527
1961
ืื‘ืœ ื–ื• ืœื ื”ื™ืชื” ื”ื“ืื’ื” ื”ื›ื™ ื’ื“ื•ืœื” ืฉืœื”ื.
06:48
Their biggest concern was trust.
127
408529
2795
ื”ื“ืื’ื” ื”ื›ื™ ื’ื“ื•ืœื” ืฉืœื”ื ื”ื™ืชื” ืืžื•ืŸ.
06:51
Now architects are liable if something goes wrong with their buildings.
128
411991
4546
ื”ืื“ืจื™ื›ืœื™ื ืื—ืจืื™ื ืœื›ืš ืฉืžืฉื”ื• ืžืฉืชื‘ืฉ ื‘ื‘ื ื™ื™ื ื™ื ืฉืœื”ื.
06:56
They could lose their license,
129
416537
1669
ื”ื ืขืœื•ืœื™ื ืœืื‘ื“ ืืช ื”ืจื™ืฉื™ื•ืŸ ืฉืœื”ื,
06:58
they could be fined, they could even go to prison.
130
418206
3211
ื”ื ืขืœื•ืœื™ื ืœื”ื™ืงื ืก, ื”ื ืขืœื•ืœื™ื ืืคื™ืœื• ืœื”ื™ื›ื ืก ืœื›ืœื.
07:01
And failures can happen in a million different ways.
131
421459
3086
ื•ื›ืฉืœื™ื ืขืœื•ืœื™ื ืœืงืจื•ืช ื‘ืžื™ืœื™ื•ืŸ ื“ืจื›ื™ื ืฉื•ื ื•ืช.
07:04
For example, exit doors that open the wrong way,
132
424545
3045
ืœื“ื•ื’ืžื”, ื“ืœืชื•ืช ื™ืฆื™ืื” ืฉื ืคืชื—ื•ืช ื‘ื›ื™ื•ื•ืŸ ื”ืœื-ื ื›ื•ืŸ,
07:07
leading to people being crushed in an evacuation crisis,
133
427632
4296
ื•ืื– ืื ืฉื™ื ื™ื™ืžื—ืฆื• ื‘ืขืช ืคื™ื ื•ื™ ื—ื™ืจื•ื,
07:11
or broken glass raining down onto pedestrians in the street
134
431928
4129
ืื• ืฉื‘ืจื™ ื–ื›ื•ื›ื™ืช ืฉื™ื™ืคื•ืœื• ืขืœ ื”ื•ืœื›ื™ ื”ืจื’ืœ ื‘ืจื—ื•ื‘
07:16
because the wind blows too hard and shatters windows.
135
436099
3837
ื›ื™ ื”ืจื•ื— ื ื•ืฉื‘ืช ื—ื–ืง ืžื“ื™ ื•ืžื ืคืฆืช ื—ืœื•ื ื•ืช.
07:20
So why would an architect trust an AI system with their job,
136
440311
3921
ืื– ืœืžื” ืฉืื“ืจื™ื›ืœ ื™ืืฆื™ืœ ืืช ืขื‘ื•ื“ืชื• ืœืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช,
07:24
with their literal freedom,
137
444273
2670
ืœื—ื•ืคืฉ ืฉื™ืฉ ืœื”, ืคืฉื•ื˜ื• ื›ืžืฉืžืขื•,
07:26
if they couldn't go in and fix a mistake if they found it?
138
446943
3628
ืื ื”ื•ื ืœื ื™ื›ื•ืœ ืœื”ืชืขืจื‘ ื•ืœืชืงืŸ ื˜ืขื•ืช ืื ื™ืžืฆื ื›ื–ืืช?
07:31
So we need to figure out these problems today, and I'll tell you why.
139
451030
4171
ืขืœื™ื ื• ืœื”ื‘ื™ืŸ ืืช ื”ื‘ืขื™ื•ืช ื”ืืœื” ื”ื™ื•ื, ื•ืื’ื™ื“ ืœื›ื ืœืžื”.
07:35
The next wave of artificial intelligence systems, called agentic AI,
140
455201
5047
ื”ื’ืœ ื”ื‘ื ืฉืœ ืžืขืจื›ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช, โ€œืื’โ€™ื ื˜ื™ืง ืื™ื™-ืื™ื™โ€œ,
07:40
is a true tipping point
141
460289
1335
ื”ื•ื ื ืงื•ื“ืช ืžืคื ื” ืืžื™ืชื™ืช
07:41
between whether or not we retain human agency,
142
461624
4213
ื‘ื™ืŸ ื–ื” ืฉื ืงืคื™ื“ ืขืœ ื ื•ื›ื—ื•ืช ืื ื•ืฉื™ืช ื•ื‘ื™ืŸ ืื ืœื,
07:45
or whether or not AI systems make our decisions for us.
143
465837
3795
ืื• ื‘ื™ืŸ ื–ื” ืฉืžืขืจื›ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื™ื—ืœื™ื˜ื• ื‘ืžืงื•ืžื ื• ืื• ืœื.
07:50
Imagine an AI agent as kind of like a personal assistant.
144
470008
3211
ื“ืžื™ื™ื ื• ื’ื•ืจื ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื›ืœืฉื”ื• ื›ืžื• ืขื•ื–ืจ ืื™ืฉื™.
07:53
So, for example, a medical agent might determine
145
473261
3712
ื›ืš, ืœืžืฉืœ, ื’ื•ืจื ืจืคื•ืื™ ืขืฉื•ื™ ืœืงื‘ื•ืข
ืื ื”ืžืฉืคื—ื” ืฉืœื›ื ื–ืงื•ืงื” ืฆืจื™ื›ื” ืœื”ื™ืคื’ืฉ ืขื ืจื•ืคื ืื• ืœื,
07:57
whether or not your family needs doctor's appointments,
146
477015
2585
07:59
it might refill prescription medications, or in case of an emergency,
147
479642
3379
ื’ื•ืจื ื›ื–ื” ืื•ืœื™ ื™ื“ืื’ ืœืžืœืื™ ืชืจื•ืคื•ืช ื”ืžืจืฉื, ืื• ื‘ืžืงืจื” ื—ื™ืจื•ื,
08:03
send medical records to the hospital.
148
483062
2586
ื™ืฉืœื— ืœื‘ื™ืช ื”ื—ื•ืœื™ื ืืช ื”ืชื™ืง ื”ืจืคื•ืื™.
08:05
But AI agents can't and won't exist
149
485690
2836
ืื‘ืœ ื’ื•ืจืžื™ ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืœื ื™ื›ืœื• ืœื”ืชืงื™ื™ื ื•ืœื ื™ืชืงื™ื™ืžื•
08:08
unless we have a true right to repair.
150
488568
2210
ืืœื ืื ืชื”ื™ื” ืœื ื• ื–ื›ื•ืช ืืžื™ืชื™ืช ืœืชืงืŸ.
08:10
What parent would trust their child's health to an AI system
151
490820
4922
ืื™ื–ื” ื”ื•ืจื” ื™ืคืงื™ื“ ืืช ื‘ืจื™ืื•ืช ื™ืœื“ื™ื• ื‘ื™ื“ื™ ืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช
08:15
unless you could run some basic diagnostics?
152
495783
2795
ื‘ืœื™ ืฉื™ื•ื›ืœ ืœื”ืจื™ืฅ ืื™ื–ื” ืื‘ื—ื•ืŸ ื‘ืกื™ืกื™?
08:18
What professional would trust an AI system with job decisions,
153
498619
4213
ืื™ื–ื” ืื™ืฉ ืžืงืฆื•ืข ื™ืคืงื™ื“ ืื™ื•ืฉ ืžืฉืจื•ืช ื‘ื™ื“ื™ ืžืขืจื›ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช,
08:22
unless you could retrain it the way you might a junior employee?
154
502874
4296
ืืœื ืื ื›ืŸ ื™ื•ื›ืœ ืœื”ื›ืฉื™ืจ ืื•ืชื” ืžื—ื“ืฉ ื›ืžื• ืฉืขื•ืฉื™ื ืขื ืขื•ื‘ื“ ื–ื•ื˜ืจ?
08:28
Now, a right to repair might look something like this.
155
508129
2878
ื”ื–ื›ื•ืช ืœืชืงืŸ ืขืฉื•ื™ื” ืœื”ื™ืจืื•ืช ื‘ืขืจืš ื›ืš.
08:31
You could have a diagnostics board
156
511007
2210
ื™ื›ื•ืœ ืœื”ื™ื•ืช ืœื›ื ืœื•ื— ืื‘ื—ื•ืŸ
08:33
where you run basic tests that you design,
157
513259
3003
ืฉื‘ื• ืชื‘ืฆืขื• ื‘ื“ื™ืงื•ืช ื‘ืกื™ืกื™ื•ืช ืœืคื™ ืชื›ื ื•ืŸ ืฉืœื›ื,
08:36
and if something's wrong, you could report it to the company
158
516304
2836
ื•ืื ืžืฉื”ื• ืœื ื‘ืกื“ืจ, ืชื•ื›ืœื• ืœื“ื•ื•ื— ืขืœ ื›ืš ืœื—ื‘ืจื”
ื•ืœืงื‘ืœ ืชืฉื•ื‘ื” ื›ืฉื–ื” ืžืชื•ืงืŸ.
08:39
and hear back when it's fixed.
159
519140
1585
08:40
Or you could work with third parties like ethical hackers
160
520725
3128
ืื• ืฉืชื•ื›ืœื• ืœืขื‘ื•ื“ ืขื ืฆื“ ืฉืœื™ืฉื™ ื›ืžื• ืคืฆื—ื ื™ื ืืชื™ื™ื
08:43
who make patches for systems like we do today.
161
523895
2586
ืฉืžื›ื™ื ื™ื ื˜ืœืื™ื ืœืžืขืจื›ื•ืช ื›ืžื• ืฉืื ื• ืขื•ืฉื™ื ื”ื™ื•ื.
08:46
You can download them and use them to improve your system
162
526481
2711
ืชื•ื›ืœื• ืœื”ื•ืจื™ื“ ืื•ืชื ื•ืœืฉืคืจ ื‘ืขื–ืจืชื ืืช ื”ืžืขืจื›ืช ืฉืœื›ื
ื›ืคื™ ืฉืืชื ืจื•ืฆื™ื ืฉื”ื™ื ืชืฉืชืคืจ.
08:49
the way you want it to be improved.
163
529233
1919
08:51
Or you could be like these intrepid farmers and learn to program
164
531194
3628
ืื• ืฉืชื•ื›ืœื• ืœื”ื™ื•ืช ื›ืžื• ื”ื—ืงืœืื™ื ื”ืืžื™ืฆื™ื ื”ืืœื”,
ื•ืœืœืžื•ื“ ืœืชื›ื ืช ื•ืœื›ื™ื™ืœ ืืช ื”ืžืขืจื›ื•ืช ืฉืœื›ื.
08:54
and fine-tune your own systems.
165
534864
3003
08:58
We won't achieve the promised benefits of artificial intelligence
166
538618
4171
ืœื ื ืฉื™ื’ ืืช ื”ื™ืชืจื•ื ื•ืช ื”ืžื•ื‘ื˜ื—ื™ื ืฉืœ ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช
09:02
unless we figure out how to bring people into the development process.
167
542789
5046
ืื ืœื ื ื‘ื™ืŸ ื›ื™ืฆื“ ืœืฉืœื‘ ืื ืฉื™ื ื‘ืชื”ืœื™ืš ื”ืคื™ืชื•ื—.
09:08
I've dedicated my career to responsible AI,
168
548377
3504
ื”ืงื“ืฉืชื™ ืืช ื”ืงืจื™ื™ืจื” ืฉืœื™ ืœื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืื—ืจืื™ืช,
09:11
and in that field we ask the question,
169
551923
3253
ื•ื‘ืชื—ื•ื ื–ื” ืื ื• ืฉื•ืืœื™ื ืืช ื”ืฉืืœื”,
09:15
what can companies build to ensure that people trust AI?
170
555218
4963
ืžื” ื”ื—ื‘ืจื•ืช ื™ื›ื•ืœื•ืช ืœื‘ื ื•ืช ื›ื“ื™ ืœื”ื‘ื˜ื™ื— ืฉืื ืฉื™ื ื™ืกืžื›ื• ืขืœ ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช?
09:20
Now, through these red-teaming exercises, and by talking to you,
171
560723
3837
ื‘ืขื–ืจืช ืชืจื’ื™ืœื™ ืฆื•ื•ืช ืื“ื•ื ืืœื”, ื•ื”ืฉื™ื—ื•ืช ืื™ืชื›ื,
09:24
I've come to realize that we've been asking the wrong question all along.
172
564602
5339
ื”ื‘ื ืชื™ ืฉืฉืืœื ื• ื›ืœ ื”ื–ืžืŸ ืืช ื”ืฉืืœื” ื”ืœื-ื ื›ื•ื ื”.
09:30
What we should have been asking is what tools can we build
173
570566
3921
ืžื” ืฉื”ื™ื™ื ื• ืฆืจื™ื›ื™ื ืœืฉืื•ืœ ื”ื•ื ืื™ืœื• ื›ืœื™ื ื ื•ื›ืœ ืœื‘ื ื•ืช
09:34
so people can make AI beneficial for them?
174
574529
4045
ื›ื“ื™ ืฉืื ืฉื™ื ื™ื•ื›ืœื• ืœื”ืคื•ืš ืืช ื”ื‘ื™ื ื” ื”ืžืœืื›ื•ืชื™ืช ืœืžื•ืขื™ืœื” ืขื‘ื•ืจื?
09:39
Technologists can't do it alone.
175
579117
2460
ืื ืฉื™ ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ืœื ื™ื›ื•ืœื™ื ืœืขืฉื•ืช ืืช ื–ื” ื‘ืขืฆืžื.
09:41
We can only do it with you.
176
581619
2419
ื ื•ื›ืœ ืœืขืฉื•ืช ืืช ื–ื” ืืš ื•ืจืง ื™ื—ื“ ืื™ืชื›ื.
09:44
Thank you.
177
584080
1168
ืชื•ื“ื” ืœื›ื.
09:45
(Applause)
178
585248
2794
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7